Details
| Original language | English |
|---|---|
| Pages (from-to) | 3048-3053 |
| Number of pages | 6 |
| Journal | IEEE Control Systems Letters |
| Volume | 7 |
| Publication status | Published - 7 Jul 2023 |
| Externally published | Yes |
Abstract
Many application fields, e.g., robotic surgery, autonomous piloting, and wearable robotics greatly benefit from advances in robotics and automation. A common task is to control an unknown nonlinear system such that its output tracks a desired reference signal for a finite duration of time. A learning control method that automatically and efficiently designs output feedback controllers for this task would greatly boost practicality over time-consuming and labour-intensive manual system identification and controller design methods. In this contribution we propose Automatic Neural Ordinary Differential Equation Control (ANODEC), a data-efficient automatic design of output feedback controllers for finite-time reference tracking in systems with unknown nonlinear dynamics. In a two-step approach, ANODEC first identifies a neural ODE model of the system dynamics from input-output data of the system dynamics and then exploits this data-driven model to learn a neural ODE feedback controller, while requiring no knowledge of the actual system state or its dimensionality. In-silico validation shows that ANODEC is able to - automatically - design competitive controllers that outperform two controller baselines, and achieves an on average ≈ 30 % / 17 % lower median RMSE. This is demonstrated in four different nonlinear systems using multiple, qualitatively different and even out-of-training-distribution reference signals.
Keywords
- Autonomous systems, data-driven modeling, learning systems, motion control, neural networks
ASJC Scopus subject areas
- Engineering(all)
- Control and Systems Engineering
- Mathematics(all)
- Control and Optimization
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: IEEE Control Systems Letters, Vol. 7, 07.07.2023, p. 3048-3053.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - Neural ODEs for Data-Driven Automatic Self-Design of Finite-Time Output Feedback Control for Unknown Nonlinear Dynamics
AU - Bachhuber, Simon
AU - Weygers, Ive
AU - Seel, Thomas
PY - 2023/7/7
Y1 - 2023/7/7
N2 - Many application fields, e.g., robotic surgery, autonomous piloting, and wearable robotics greatly benefit from advances in robotics and automation. A common task is to control an unknown nonlinear system such that its output tracks a desired reference signal for a finite duration of time. A learning control method that automatically and efficiently designs output feedback controllers for this task would greatly boost practicality over time-consuming and labour-intensive manual system identification and controller design methods. In this contribution we propose Automatic Neural Ordinary Differential Equation Control (ANODEC), a data-efficient automatic design of output feedback controllers for finite-time reference tracking in systems with unknown nonlinear dynamics. In a two-step approach, ANODEC first identifies a neural ODE model of the system dynamics from input-output data of the system dynamics and then exploits this data-driven model to learn a neural ODE feedback controller, while requiring no knowledge of the actual system state or its dimensionality. In-silico validation shows that ANODEC is able to - automatically - design competitive controllers that outperform two controller baselines, and achieves an on average ≈ 30 % / 17 % lower median RMSE. This is demonstrated in four different nonlinear systems using multiple, qualitatively different and even out-of-training-distribution reference signals.
AB - Many application fields, e.g., robotic surgery, autonomous piloting, and wearable robotics greatly benefit from advances in robotics and automation. A common task is to control an unknown nonlinear system such that its output tracks a desired reference signal for a finite duration of time. A learning control method that automatically and efficiently designs output feedback controllers for this task would greatly boost practicality over time-consuming and labour-intensive manual system identification and controller design methods. In this contribution we propose Automatic Neural Ordinary Differential Equation Control (ANODEC), a data-efficient automatic design of output feedback controllers for finite-time reference tracking in systems with unknown nonlinear dynamics. In a two-step approach, ANODEC first identifies a neural ODE model of the system dynamics from input-output data of the system dynamics and then exploits this data-driven model to learn a neural ODE feedback controller, while requiring no knowledge of the actual system state or its dimensionality. In-silico validation shows that ANODEC is able to - automatically - design competitive controllers that outperform two controller baselines, and achieves an on average ≈ 30 % / 17 % lower median RMSE. This is demonstrated in four different nonlinear systems using multiple, qualitatively different and even out-of-training-distribution reference signals.
KW - Autonomous systems
KW - data-driven modeling
KW - learning systems
KW - motion control
KW - neural networks
UR - http://www.scopus.com/inward/record.url?scp=85164403404&partnerID=8YFLogxK
U2 - 10.1109/LCSYS.2023.3293277
DO - 10.1109/LCSYS.2023.3293277
M3 - Article
AN - SCOPUS:85164403404
VL - 7
SP - 3048
EP - 3053
JO - IEEE Control Systems Letters
JF - IEEE Control Systems Letters
ER -